Is ChatGPT an existential threat to accountants or an aid?

Much of accounting and tax compliance involves repetition and follows predictable patterns. Large Language Models, such as ChatGPT, are trained through countless repetitions leading to superhuman pattern recognition. 

Nicolas Boucher, a LinkedIn FP&A and efficiency influencer, and experts from PwC South Africa share their insights on whether ChatGPT heralds the decline of the accountant or whether it will free accountants from low-level tasks. 

If you fear ChatGPT and AI, you’re in good company. 

In March, Elon Musk, Apple co-founder Steve Wozniak, leading AI experts, and over 1 000 other notable persons signed a letter calling for a six-month pause in giant AI experiments of models more potent than ChatGPT 4.

The letter lists various reasons to be wary of AI developments, “Contemporary AI systems are now becoming human-competitive at general tasks, and we must ask ourselves: Should we let machines flood our information channels with propaganda and untruth? Should we automate away all the jobs, including the fulfilling ones? Should we develop nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us?” 

A recent report by Goldman Sachs estimates that 15 to 35 percent of work will be exposed to automation. Unlike other technological developments, manual labourers are reasonably safe. Instead, it’s office workers who will be hit hardest. 

SARS estimates that there are 25 000 tax practitioners in South Africa. If AI continues to make tax compliance more efficient, how many will the country need in a decade? 

There are two ways of thinking of about this, says Nicolas Boucher, a German-based finance professional and LinkedIn influencer with nearly 380 000 followers. The truth is likely somewhere in the middle. 

“I had a discussion yesterday with auditors and with tax people, and we are convinced the companies who are going to embrace it will have so much work because people will come to them, they will know it’s cheaper, and they can get more value than their actual tax expert.

“If you work in this [kind of] company, your job is secure. If you work in a company where you don’t embrace it and you don’t provide more value than a computer, then you are at risk.”

How to future-proof your accounting job

Boucher says there are two options. 

One option is to specialise in an area of accounting that computers cannot do very well either because of the complexity or because it requires you to have a deep professional network.

Mohil Subban, emerging technology consulting lead at PwC South Africa, points out, “Large Language Models/GPT models may not be able to execute specialised tax review or advisory functions. These will most likely be left to specialists to handle. AI models may be able to help accelerate the first draft, but we feel that specialist review may always need to be performed at least whilst this technology is maturing.”

Boucher says, “The other type of job that will be more and more important is being the tax consultant who has enough knowledge and experience and technological understanding to ask and to parameter ChatGPT and all of the other tools to do the work ten times faster than what he was doing before.” 

“And when the output is there, having the professional scepticism and experience to review the work and approve or correct it.”

“AI is not a magic wand.”

Professional scepticism is a term Boucher comes back to repeatedly.  

“I think it’s important that you highlight that ChatGPT is not the magic wand for everything,” he says. 

“If I ask somebody in my team to analyse figures and they give me the answer in an email with some absurd information inside and I come back to them and say, ‘Well, are you aware that this doesn’t make sense?’ And they don’t own the mistake and say, ‘Well, that was just from ChatGPT, it’s not me’. Then I will answer, ‘Okay, then I don’t need you. I can also type into GPT and get this answer’.”

Riaan Singh, experience consulting lead at PwC South Africa says, “Our advice would be to treat GPT model interactions in a similar way to the results of an internet search - you should verify the output and not fully rely on what has been generated.”

PwC experts flag other areas of concern. Singh says, “We also advise not to share any sensitive or confidential information with publicly available GPT models/APIs that connect to them.” 

Subban says, “We do believe that while there may be massive impacts felt by many different industries, we caution users to be wary of relying too heavily on this progressive technology as it still needs to mature and be industrialised, secured, tested and sanctioned by the right people so it can do the right thing at the right time, in the right context and in the right way.”

ChatGPT and AI’s near future accounting applications

ChatGPT applications may still need to be matured, but many, including Boucher, are working on understanding how AI can help finance professionals.

The greatest efficiency gains – and potential job disruptions – may be a little way off. Mark Allderman, technology consulting lead at PwC says, “The next phase of adoption includes deeper training and people using GPT models to do things that existing methods take longer to do or do not produce as complete of a result, bringing forth major time efficiencies. We believe it is at this point, that we feel GPT models will be massively adopted in tasks we perform regularly.”

Boucher gives the example of a financial statement analysis using a vanilla version of ChatGPT which may provide inaccurate or incomplete information. A user can mitigate this by providing very specific prompting to give ChatGPT context and essentially train it, but this is time-consuming and can be done using Excel more efficiently.

However, increasingly products are in development that feed accounting specific information, context and instructions to ChatGPT or other AI using APIs.

Boucher says he has seen examples of this emerging from Financial Planning and Analysis vendors and accounting cloud software providers. “When I ask it to calculate the cash conversion cycle, it will know exactly which accounts to take, make the calculation and feed the result.”

How can accountants use ChatGPT now?

Boucher believes AI is particularly suited to tasks accountants stereotypically struggle with. “Most of the time, accountants went into accounting because they love the numbers, but less so writing.” One of the examples Boucher references is of accounts receivables correspondence. 

“It’s really demanding to go behind every client and ask them to pay you and to contact them,” says Boucher.  

“You could ask ChatGPT to draft emails with a specific tone,” says Boucher. You could also ask. ChatGPT to give you three levels of severity. The first one, nice. The second one is already a good reminder. And the third one is really severe. Basically the last reminder before legal action.”

This application is handy for small companies. “It’s really helpful because it gives them a template," says Boucher. "It gives them emails they can write and then reuse.”

Boucher believes this application is also relevant for companies that already have templates. “I had a conversation with people working for companies where they say, ‘well, I cannot always use my template because sometimes there is some specific situation. Then I have to change the template myself, and it takes me half an hour instead of five minutes.”

In this instance, ChatGPT can likely generate a unique email copy, with adjusted context, faster, particularly if English is not your first language. 

“It’s a great application when you work in an international environment and it’s something that can be used on any team, not just finance. But in finance, we are not good with words. So we welcome this capability,” says Boucher.

Are tax practices in 2020 in the same position as newspapers in 2000?

In a few years, who will be left invoicing and how much of that work will be done by humans as opposed to AI, is an open question. 

Optimists note that new technology also leads to new jobs. Although those displaced and not necessarily the same people suited to the new jobs. Not every tax clerk or practitioner can become an AI-assisted corporate tax advisor.    

The Goldman Sachs report notes, “Technological change displaced workers and created new employment opportunities at roughly the same rate for the first half of the post-war period, but has displaced workers at a faster pace than it has created new opportunities since the 1980s.

“These results suggest that the direct effects of generative AI on labour demand could be negative in the near-term if AI affects the labour market in a manner similar to earlier advances in information technology, although the effects on labour productivity growth would still be positive.”

For now, though, at least drafting emails is less awkward.